A light-weighted slope geological disaster monitoring and early warning method and system

Abstract

This invention relates to the field of geological disaster monitoring technology, and discloses a lightweight slope geological disaster monitoring and early warning method and system. The method includes acquiring real-time slope monitoring data, obtaining multi-scale influence layers, performing dynamic coupling calculations on the multi-scale influence layers to obtain time-varying coupling parameters, mapping the time-varying coupling parameters to obtain a coupling matrix, calculating risk factors using the isolated forest algorithm on the real-time slope monitoring data, processing the multi-scale influence layers, time-varying coupling strength parameters, and risk factors to obtain a slope disaster risk index, and constructing an early warning threshold adjustment mechanism to adaptively adjust the early warning threshold. This invention can improve the accuracy of disaster risk prediction, achieve lightweight local processing throughout the entire process, reduce system response latency, and use adaptive threshold adjustment to ensure that the system can adjust the early warning results in real time according to the long-term trend of the slope, solving the problems of fragmented cross-scale fusion, high latency, and weak anomaly detection in existing technologies.

Description

Technical Field

[0001] This invention relates to the field of geological disaster monitoring technology, and more specifically, to a lightweight slope geological disaster monitoring and early warning method and system. Background Technology

[0002] Monitoring and early warning of slope disasters has become a crucial link in ensuring safety and improving traffic efficiency. Current disaster early warning schemes typically have the following limitations: many systems rely solely on single data such as water pressure or land displacement for judgment, resulting in high false alarm and false negative rates, low sensitivity, a lack of collaborative monitoring through multi-level sensor networks, limited data fusion and analysis capabilities, and an inability to fully exploit the value of data. Risk prediction also lacks advanced algorithm support, leading to insufficient prediction accuracy and lead time. Furthermore, early warning models are not intelligent enough, with most existing systems using pre-set fixed thresholds to determine whether a slope poses a risk.

[0003] However, slope disasters are the result of multiple factors working together. The geological conditions, soil and rock properties, and slope morphology of different slopes vary greatly. Fixed threshold judgment methods lack scientific rigor and cannot adapt to the actual conditions of different slopes, nor can they provide early warnings based on the natural environment and practical realities. Therefore, there is an urgent need for a disaster early warning system that can integrate multiple factors, cope with complex environmental changes and the slope's own creep process, and adjust accordingly. This system would enable accurate prediction and proactive control of slope disasters, improving slope safety and maintenance efficiency.

[0004] No effective solutions have yet been proposed to address the problems in the relevant technologies. Summary of the Invention

[0005] To address the problems in related technologies, this invention proposes a lightweight slope geological disaster monitoring and early warning method and system to overcome the aforementioned technical problems existing in the current related technologies.

[0006] Therefore, the specific technical solution adopted by the present invention is as follows: According to one aspect of the present invention, a lightweight slope geological disaster monitoring and early warning method is provided, comprising: S1. Standardize the pre-acquired real-time slope monitoring data and decompose the processed slope monitoring data to obtain a multi-scale influence layer. S2. Using the constructed third-order tensor coupling model, dynamic coupling calculations are performed on the multi-scale influence layers to obtain time-varying coupling parameters. The time-varying coupling parameters are then mapped to obtain a coupling matrix, which is used to reflect the degree of correlation between the multi-scale influence layers. S3. Using the standardized real-time slope monitoring data as a sample set, the isolated forest algorithm is used to calculate the risk factors from the real-time slope monitoring data. S4. The multi-scale influence layer, time-varying coupling strength parameters and risk factors are processed to obtain the input vector. The input vector is then calculated using a gated cyclic unit network to obtain the slope disaster risk index. S5. Based on the slope disaster risk index and combined with historical data of slope monitoring data, a warning threshold adjustment mechanism is constructed to adaptively adjust the warning threshold.

[0007] Preferably, the standardization process of the pre-acquired real-time slope monitoring data and the decomposition of the processed slope monitoring data to obtain a multi-scale influence layer include: S11. Real-time slope monitoring data includes surface displacement data, soil tilt data, water pressure data, and rainfall data; S12. The real-time slope monitoring data is processed using the Z-Score standardization method to obtain a unified time-series dataset; S13. Perform time-frequency decomposition on the unified time series dataset to obtain a multi-scale influence layer, which includes a slow layer, a medium-speed layer, and a fast layer.

[0008] Preferably, the step of using the constructed third-order tensor coupling model to perform dynamic coupling calculations on the multi-scale influence layer to obtain time-varying coupling parameters, and mapping the time-varying coupling parameters to obtain the coupling matrix includes: S21. Construct a time window using a preset time length, obtain vector data of the multi-scale influence layer in the same time window, and use the vector data as the calculation benchmark. S22. Within the time window, for any two influence layers in the multi-scale influence layer, calculate the covariance of their vector data and obtain the covariance value. S23. Based on the preset attenuation coefficient, the time-varying coupling parameters between the corresponding influence layers are calculated by combining the covariance value.

[0009] Preferably, within the time window, for any two influence layers in the multi-scale influence layers, calculating the covariance of their vector data to obtain the covariance value includes: When the covariance is positive, any two influencing layers have a positive cooperative relationship; when the covariance is negative, any two influencing layers have a negative cooperative relationship.

[0010] Preferably, the standardized real-time slope monitoring data is used as a sample set, and the isolated forest algorithm is used to calculate the risk factors from the real-time slope monitoring data, including: S31. Obtain standardized slope detection data, construct a historical sample set, use the historical sample set to construct an isolation tree, and extract multiple feature dimensions for each sample point in the historical sample set. Among these, several characteristic latitudes include displacement, tilt, water pressure, and rainfall; S32. Obtain the maximum and minimum values ​​of at least one feature dimension of the sample point, determine a split value between the maximum and minimum values, divide the sample point into left and right subtrees according to the split value, and repeat the division process until each sample point is isolated to a preset value. S33. Select at least one sample point from the historical sample set, and use the average length of the sample point to each isolation tree as the baseline length. S34. Acquire the sample to be tested in real time, traverse each isolation tree for the sample to be tested, and calculate the average length of the sample point in each isolation tree. S35. Calculate the average length of the sample points in the sample to be tested and the baseline length to obtain the probability value of abnormal sample points, and use this probability value as a risk factor.

[0011] Preferably, the process of processing the multi-scale influence layer, time-varying coupling strength parameters, and risk factors to obtain an input vector, and then using a gated recurrent unit network to calculate the slope disaster risk index, includes: S41. The multi-scale influence layer, time-varying coupling strength parameters, and risk factors are concatenated to form the input vector at the current moment. S42. Input the current input vector into the gated recurrent unit network, and calculate the candidate hidden state and the hidden state through the reset gate and the update gate; S43. Calculate the hidden state using the fully connected layer to obtain the slope disaster risk index.

[0012] Preferably, the step of inputting the current input vector into the gated recurrent unit network and calculating the candidate hidden state and the hidden state through the reset gate and the update gate includes: S421. The previous hidden state is processed element-wise using the reset gate. The processed previous hidden state is concatenated with the current input vector to obtain a fused feature vector. The fused feature vector is then non-linearly mapped using an activation function to obtain the candidate hidden state. S422. The hidden state and candidate hidden state at the previous time step of the update gate are weighted and processed, and the element-wise weighted summation is used to calculate the result.

[0013] Preferably, the expression for the multi-scale influence layer is: ; ; ; In the formula, Indicates a fast layer. Indicates medium speed layer ， Indicates a slow layer. G Represents surface displacement data. I Indicates soil tilt data, Represents rainfall data, Represents water pressure data. Indicates the wavelet scale. Represented as mother wavelet, Indicates the time shift amount. This indicates the rainfall over 24 hours. , and Both represent attenuation coefficients. This represents the change in water pressure over 24 hours. Indicates the long-term trend term. Indicates periodic fluctuations. This represents random noise and / or residuals.

[0014] Preferably, the expression for the time-varying coupling parameter is: ; In the formula, and This represents the multi-scale influence layer, with values ​​ranging from [value range missing]. , ， , t Indicates a time window. This represents the impact layer data within the time window. Indicates the time window and The covariance operator, Indicates the time window The mean, Indicates the time window The mean, This represents the tensor decay factor between different influence layers. This represents the attenuation coefficient between multi-scale influence layers.

[0015] According to another aspect of the present invention, a lightweight slope geological disaster monitoring and early warning system is provided, comprising: The scale decomposition module is used to standardize the pre-acquired real-time slope monitoring data and decompose the processed slope monitoring data to obtain multi-scale influence layers. The coupling calculation module is used to utilize the constructed third-order tensor coupling model to perform dynamic coupling calculations on the multi-scale influence layers, obtain time-varying coupling parameters, and map the time-varying coupling parameters to obtain a coupling matrix, which is used to reflect the degree of correlation between the multi-scale influence layers. The risk calculation module is used to take the standardized real-time slope monitoring data as a sample set and use the isolated forest algorithm to calculate the risk factors from the real-time slope monitoring data. The index calculation module is used to process the multi-scale influence layer, time-varying coupling strength parameters and risk factors to obtain the input vector. The input vector is then calculated using a gated cyclic unit network to obtain the slope disaster risk index. The adaptive adjustment module is used to construct an early warning threshold adjustment mechanism based on the slope disaster risk index and combined with historical data of slope monitoring data, and to adaptively adjust the early warning threshold.

[0016] The beneficial effects of this invention are as follows: By combining multiple sensors such as inclinometers, water pressure sensors, GNSS sensors, and rain gauges, it fully utilizes the advantages of multi-source data. By integrating the outputs of various influencing layers using a gating mechanism, it overcomes the shortcomings of a single data source, improving the accuracy of disaster risk prediction. The entire process is processed locally and lightweight, reducing system response latency and ensuring timely disaster emergency response. Adaptive threshold adjustment ensures that the system can adjust the early warning results in real time according to the long-term trend of the slope, exhibiting good dynamic response capabilities. This avoids the problem of traditional methods over-relying on fixed thresholds, which leads to inaccurate slope disaster early warnings due to changes in the geological environment. It is suitable for rapid early warning of various types of lithological slopes, solving the problems of fragmented cross-scale fusion, high latency, and weak anomaly detection in existing technologies. Attached Figure Description

[0017] To more clearly illustrate the technical solutions in the embodiments of the present invention or the prior art, the drawings used in the embodiments will be briefly introduced below. Obviously, the drawings described below are only some embodiments of the present invention. For those skilled in the art, other drawings can be obtained based on these drawings without creative effort.

[0018] Figure 1 This is a flowchart of a lightweight slope geological disaster monitoring and early warning method according to an embodiment of the present invention; Figure 2 This is a schematic diagram of a lightweight slope geological disaster monitoring and early warning system according to an embodiment of the present invention; Figure 3 This is a schematic diagram of the sensor installation section in a lightweight slope geological disaster monitoring and early warning method according to an embodiment of the present invention; Figure 4This is one of the schematic diagrams of slope monitoring data in a lightweight slope geological disaster monitoring and early warning method according to an embodiment of the present invention; Figure 5 This is the second schematic diagram of slope monitoring data in a lightweight slope geological disaster monitoring and early warning method according to an embodiment of the present invention; Figure 6 This is the third schematic diagram of slope monitoring data in a lightweight slope geological disaster monitoring and early warning method according to an embodiment of the present invention; Figure 7 This is the fourth schematic diagram of slope monitoring data in a lightweight slope geological disaster monitoring and early warning method according to an embodiment of the present invention; Figure 8 This is a schematic diagram of the coupling matrix relationship in a lightweight slope geological disaster monitoring and early warning method according to an embodiment of the present invention; Figure 9 This is a schematic diagram of risk factor detection in a lightweight slope geological disaster monitoring and early warning method according to an embodiment of the present invention; Figure 10 This is a schematic diagram illustrating the dynamic adjustment of the early warning threshold in a lightweight slope geological disaster monitoring and early warning method according to an embodiment of the present invention.

[0019] In the picture: 1. Scale decomposition module; 2. Coupled calculation module; 3. Risk calculation module; 4. Index calculation module; 5. Adaptive adjustment module. Detailed Implementation

[0020] To further illustrate the various embodiments, the present invention provides accompanying drawings, which are part of the disclosure of the present invention. These drawings are mainly used to illustrate the embodiments and can be used in conjunction with the relevant descriptions in the specification to explain the operating principles of the embodiments. With reference to these drawings, those skilled in the art should be able to understand other possible implementation methods and the advantages of the present invention. The components in the drawings are not drawn to scale, and similar component symbols are generally used to represent similar components.

[0021] According to an embodiment of the present invention, a lightweight slope geological disaster monitoring and early warning method and system are provided.

[0022] The present invention will now be further described in conjunction with the accompanying drawings and specific embodiments, such as... Figure 1 As shown in one embodiment of the present invention, a method for monitoring and early warning of geological hazards on lightweight slopes is provided, comprising: S1. Standardize the pre-acquired real-time slope monitoring data and decompose the processed slope monitoring data to obtain a multi-scale influence layer. As a preferred embodiment, the standardization of the pre-acquired real-time slope monitoring data and the decomposition of the processed slope monitoring data to obtain a multi-scale influence layer include the following steps: the real-time slope monitoring data includes surface displacement data, soil tilt data, water pressure data, and rainfall data; the real-time slope monitoring data is processed using the Z-Score standardization method to obtain a unified time-series dataset; the unified time-series dataset is decomposed into time and frequency to obtain a multi-scale influence layer, wherein the multi-scale influence layer includes a slow layer, a medium-speed layer, and a fast layer.

[0023] Specifically, this invention utilizes GNSS (Global Navigation Satellite System), inclinometers, rain gauges, and water pressure gauges to collect slope monitoring data in real time, such as... Figure 3 As shown, four points were selected on the slope, and GNSS was installed at each point. At the same time, stainless steel monitoring rods were drilled and buried, and servo inclinometers and water pressure gauges were set up at intervals. A GNSS displacement observation station was set up at each point.

[0024] The IN01 connecting rod is 30m deep in total. A rain gauge (IN01-JY) is installed on the ground, and the first inclinometer (IN01-CX01) is installed 1m below the ground. After that, an inclinometer (IN01-CX02 to IN01-CX12) is installed every 3m. A water pressure gauge (IN01-SY01) is installed at the bottom of the connecting rod (30m below the ground). The IN02 connecting rod is 10m deep in total. The first inclinometer (IN02-CX01) is installed 2m deep underground. After that, an inclinometer is installed every 2m (IN02-CX02 to IN02-CX06). A water pressure gauge (IN02-SY01) is installed at the bottom of the connecting rod (10m deep underground). The IN03 link is 30m deep in total. A rain gauge (IN03-JY) is installed on the ground. The inclinometer is arranged in the same way as the IN01 link (IN03-CX01 to IN03-CX12). A water pressure gauge (IN03-SY01 to IN03-SY03) is installed at depths of 10m, 20m and 30m underground. The IN04 connecting rod is 15m deep in total. A rain gauge (IN04-JY) is installed on the ground, the first inclinometer (IN04-CX01) is installed 1m underground, and then an inclinometer (IN04-CX02 to IN04-CX07) is installed every 2m. A water pressure gauge (IN04-SY01) is installed at the bottom of the connecting rod (15m underground).

[0025] Among them, GNSS and inclinometers accurately monitor surface displacement, acquire real-time dynamic changes in soil tilt angle and deep displacement data, water pressure gauges measure pore water pressure, monitor groundwater level changes in real time, and acquire water pressure data, while rain gauges are used to monitor rainfall intensity and accumulation, and acquire rainfall intensity and accumulation values.

[0026] Next, as Figures 4 to 7 As shown, the above sensors are connected to the data acquisition instrument to obtain slope monitoring data, and the slope monitoring data is processed using the Z-Score standardization method to convert it into dimensionless standardized data with a mean of 0 and a standard deviation of 1.

[0027] The expression for standardized data is: ; In the formula, Let i represent standardized data, where i∈(1,2,3,4). This indicates slope monitoring data. This represents the mean of the data. The standard deviation of the data.

[0028] Next, the standardized slope monitoring data was decomposed to obtain a multi-scale influence layer containing a slow layer, a medium-speed layer, and a fast layer. The fast layer corresponds to the influence of surface displacement on the slope, the medium-speed layer corresponds to the influence of rainfall and pore pressure on the slope, and the slow layer corresponds to the influence of soil tilting and movement on the slope.

[0029] The expression for the fast layer is: ; In the formula, G Represents surface displacement data. Indicates the wavelet scale. Represented as mother wavelet, This indicates the amount of time shift.

[0030] The expression for the slow layer is: ; In the formula, I Indicates soil tilt data, Represents rainfall data, Represents water pressure data. This indicates the rainfall over 24 hours. , and Both represent the attenuation coefficients within a multi-scale influence layer, used to describe the attenuation characteristics of physical quantities within their respective single influence layers over time. It represents the change in water pressure over 24 hours.

[0031] And the expression for the intermediate speed layer is: ; In the formula, Indicates the long-term trend term. Indicates periodic fluctuations. This represents random noise and / or residuals.

[0032] S2. Using the constructed third-order tensor coupling model, dynamic coupling calculations are performed on the multi-scale influence layers to obtain time-varying coupling parameters. The time-varying coupling parameters are then mapped to obtain a coupling matrix, which is used to reflect the degree of correlation between the multi-scale influence layers. In a preferred embodiment, the method of using the constructed third-order tensor coupling model to perform dynamic coupling calculations on multi-scale influence layers to obtain time-varying coupling parameters, and mapping these time-varying coupling parameters to obtain a coupling matrix to reflect the degree of correlation between multi-scale influence layers, includes the following steps: constructing a time window using a preset time length, acquiring vector data of the multi-scale influence layers within the same time window, and using this vector data as the calculation benchmark; within the time window, calculating the covariance of the vector data of any two influence layers in the multi-scale influence layers, and obtaining the covariance value; when the covariance value is positive, any two influence layers have a positive cooperative change relationship, and when the covariance value is negative, any two influence layers have a negative cooperative change relationship. Based on a preset attenuation coefficient, the time-varying coupling parameters between the corresponding influence layers are calculated in combination with the covariance value.

[0033] Specifically, such as Figure 8 As shown, the dynamic interaction and synergistic effect between multi-scale influence layers are quantified using a coupling matrix, reflecting the degree of correlation between the influence layers. When different influence layers change synchronously within a certain time window, a positive covariance value indicates a positive interaction between them; conversely, a negative covariance value indicates a negative interaction. The attenuation coefficient simulates the characteristic that the coupling strength between influence layers at different time scales weakens with increasing distance. The expression for the time-varying coupling parameter is: ; In the formula, and This represents the multi-scale influence layer, with values ​​ranging from [value range missing]. , ， , t Indicates a time window. This represents the impact layer data within the time window. Indicates the time window and The covariance operator, Indicates the time window The mean, Indicates the time window The mean, This represents the tensor decay factor between different influence layers. k It represents the coupling attenuation coefficient between multi-scale influence layers, used to characterize the characteristic that the interaction strength between different influence layers gradually weakens as the time window or scale difference increases, and the object of action is the coupling relationship between different influence layers.

[0034] S3. Using the standardized real-time slope monitoring data as a sample set, the isolated forest algorithm is used to calculate the risk factors from the real-time slope monitoring data. In a preferred embodiment, the step of using standardized real-time slope monitoring data as a sample set and calculating risk factors using the isolated forest algorithm includes the following steps: acquiring standardized slope monitoring data, constructing a historical sample set, using the historical sample set to construct an isolation tree, and extracting multiple feature dimensions for each sample point in the historical sample set, wherein the multiple feature dimensions include displacement, tilt, water pressure, and rainfall; obtaining the maximum and minimum values ​​of at least one feature dimension in the sample point, determining a segmentation value between the maximum and minimum values, dividing the sample point into left and right subtrees according to the segmentation value, and repeating this division process until each sample point is isolated to a preset value; selecting at least one sample point from the historical sample set, and using the average length of the sample point to each isolation tree as the baseline length; acquiring the sample to be detected in real time, traversing each isolation tree for the sample to be detected, and calculating the average length of the sample points in the sample to be detected in each isolation tree; calculating the probability value of abnormal sample points by comparing the average length of the sample points in the sample to be detected with the baseline length, and using this probability value as the risk factor.

[0035] Specifically, geological disasters often have certain warning signs. Signals such as minute ground displacements, groundwater level fluctuations, and soil position changes frequently exhibit abnormal deviations, which are all precursory characteristics and key indicators for disaster prediction. Therefore, the unsupervised learning-based isolated forest algorithm is used to detect abnormal precursors, such as... Figure 9 As shown, outlier sample points typically have shorter path lengths. Therefore, the path length can be used to determine whether a sample is outlier. The formula for calculating the probability of an outlier sample point is: ; In the formula, Indicates risk factors, n Represents the total sample size. This represents the path length from the sample point to the decision point. This represents the average path length correction factor.

[0036] S4. The multi-scale influence layer, time-varying coupling strength parameters and risk factors are processed to obtain the input vector. The input vector is then calculated using a gated cyclic unit network to obtain the slope disaster risk index. In a preferred embodiment, the process of processing the multi-scale influence layer, time-varying coupling strength parameters, and risk factors to obtain an input vector, and then calculating the slope disaster risk index using a gated recurrent unit network, includes the following steps: concatenating the multi-scale influence layer, time-varying coupling strength parameters, and risk factors to form the current input vector; inputting the current input vector into the gated recurrent unit network, calculating candidate hidden states and hidden states through reset and update gates; performing element-wise operations on the previous hidden state using the reset gate, concatenating the calculated previous hidden state with the current input vector to obtain a fused feature vector, and performing nonlinear mapping on the fused feature vector using an activation function to obtain candidate hidden states; weighting the previous hidden state and candidate hidden states using the update gate, and calculating the result using element-wise weighted summation; and finally calculating the hidden states using a fully connected layer to obtain the slope disaster risk index.

[0037] Specifically, a gated recurrent unit network is used as a nonlinear reconstructor to integrate multi-scale influence layers, time-varying coupling strength parameters, and risk factors, adaptively balancing the contribution of different input vector features to the slope disaster risk index.

[0038] Candidate hidden states and hidden states are calculated using a reset gate and an update gate. The update gate controls the proportion of historical hidden states retained, determining the degree of fusion between the current input and historical hidden states. When the update gate value is close to 1, it means that the current state will retain most of the historical information. When the value is close to 0, it tends to ignore the past and focus more on the current input information. Its expression is: ; In the formula, This represents the activation function. Represents the weight parameters. Represents the input vector. This indicates the hidden state at the previous moment. This represents the bias vector of the corresponding gate structure in the gated recurrent unit network. The bias vectors used by each gate structure are consistent in dimension and are used to translate and adjust the linear combination result of the input vector and the hidden state.

[0039] The reset gate controls the update ratio of historical hidden states, that is, it controls the update ratio of historical information, representing the contribution of the previous hidden state to the candidate hidden state. The closer the reset gate value is to 0, the more it tends to forget historical information when calculating the candidate state. The expression for the reset gate is: ; In the formula, This represents the activation function. Represents the weight parameters. Represents the input vector. This indicates the hidden state at the previous moment. This represents the bias vector of the corresponding gate structure in the gated recurrent unit network. The bias vectors used by each gate structure are consistent in dimension and are used to translate and adjust the linear combination result of the input vector and the hidden state.

[0040] The candidate hidden state is fused with historical information filtered by the reset gate and the current input vector. Then, it is passed through the update gate to balance the ratio of historical state and candidate hidden state, and the hidden state at the current time step is calculated. The expression for the candidate hidden state is: ; In the formula, W represents the weight parameter. The symbol represents the reset gate, and circle (⊙) represents element-wise multiplication. This indicates the hidden state at the previous moment. Represents the input vector. This represents the bias vector of the corresponding gate structure in the gated recurrent unit network. The bias vectors used by each gate structure are consistent in dimension and are used to translate and adjust the linear combination result of the input vector and the hidden state.

[0041] The expression for the hidden state is: ; In the formula, Indicates an update to the door. This indicates the hidden state at the previous moment. ⊙ indicates a candidate hidden state, and ⊙ indicates element-wise multiplication.

[0042] Finally, the hidden states are linearly mapped using a fully connected layer, and the final slope hazard risk index is calculated, with the following expression: ; In the formula, GHR represents the slope hazard risk index, and W represents the weighting parameter. represents the hidden state, and b represents the output bias vector of the fully connected layer, which is used to correct the bias of the result after the hidden state is linearly mapped.

[0043] After obtaining the slope disaster risk index, it is compared with the preset early warning threshold to achieve early warning of geological disasters.

[0044] S5. Based on the slope disaster risk index and combined with historical data of slope monitoring data, a warning threshold adjustment mechanism is constructed to adaptively adjust the warning threshold.

[0045] Specifically, to avoid the limitations of fixed early warning thresholds, an adaptive early warning threshold adjustment mechanism is established based on real-time calculated GHR values ​​and historical slope monitoring data. This mechanism continuously monitors the distribution of the disaster risk index over a recent period, quantifies risk fluctuations, and its dynamic threshold expression is as follows: ; In the formula, This indicates the preset warning threshold. This represents the long-term trend term of historical slope monitoring data. The standardized function representing the long-term trend term. This indicates the uncertainty of the internal risk index and is used to measure its probability of occurrence.

[0046] ; In the formula, This indicates that the risk index exceeds the threshold within time period t. The probability of.

[0047] like Figure 10 As shown, based on the long-term creep trend of the slope and real-time risk fluctuations, the system will dynamically adjust the early warning threshold. K When the GHR value calculated in real time reaches a certain range, the system will trigger an alarm of the corresponding level, driving the on-site audible and visual devices to sound an alarm.

[0048] For example, when GHR < 0.6 K When the slope condition is determined to be stable, the green light on the device will remain constantly lit.

[0049] When 0.6 K ≤GHR<0.8 K If the slope is deemed to have a slight risk, the yellow light on the device will remain on, and the cloud platform will push a "caution reminder" to the on-site monitoring personnel (including GHR value and abnormal characteristics). The cloud platform will also mark the "area of ​​concern".

[0050] When 0.8 K ≤GHR< K If the slope is determined to be at a moderate risk, the device will flash a yellow light, emit intermittent beeping sounds, and send a "caution reminder" to the on-site monitoring personnel.

[0051] When GHR≥ K If the slope is deemed to be in extreme danger, the device will flash a red light and continuously emit a buzzer, sending an emergency warning to emergency command centers at all levels and activating the emergency response plan.

[0052] According to another embodiment of the invention, such as Figure 2 As shown, a lightweight slope geological hazard monitoring and early warning system is provided, including: The scale decomposition module 1 is used to standardize the pre-acquired real-time slope monitoring data and decompose the processed slope monitoring data to obtain a multi-scale influence layer. The coupling calculation module 2 is used to utilize the constructed third-order tensor coupling model to perform dynamic coupling calculations on the multi-scale influence layers, obtain time-varying coupling parameters, and map the time-varying coupling parameters to obtain a coupling matrix, which is used to reflect the degree of correlation between the multi-scale influence layers. Risk calculation module 3 is used to take the standardized real-time slope monitoring data as a sample set and use the isolated forest algorithm to calculate the risk factors from the real-time slope monitoring data. The index calculation module 4 is used to process the multi-scale influence layer, time-varying coupling intensity parameters and risk factors to obtain the input vector, and to calculate the slope disaster risk index by using a gated cyclic unit network. The adaptive adjustment module 5 is used to construct an early warning threshold adjustment mechanism based on the slope disaster risk index and combined with historical data of slope monitoring data, and to adaptively adjust the early warning threshold.

[0053] To facilitate understanding of the above technical solutions of the present invention, the working principle or operation method of the present invention in actual process will be described in detail below.

[0054] In summary, by utilizing the technical solution of this invention, and combining multiple sensors such as inclinometers, water pressure sensors, GNSS sensors, and rain gauges, the advantages of multi-source data are fully utilized. A gating mechanism is used to fuse the outputs of each influencing layer, overcoming the shortcomings of a single data source and improving the accuracy of disaster risk prediction. The algorithm is integrated into the edge computing unit, enabling lightweight local processing throughout the entire process, reducing system response latency and ensuring timely disaster emergency response. Adaptive threshold adjustment ensures that the system can adjust the early warning results in real time according to the long-term trend of the slope, exhibiting excellent dynamic response capabilities. This avoids the problem of inaccurate slope disaster early warning due to changes in the geological environment caused by over-reliance on fixed thresholds in traditional methods. It is suitable for rapid early warning of various types of lithological slopes, solving the problems of fragmented cross-scale fusion, high latency, and weak anomaly detection in existing technologies.

[0055] Those skilled in the art will understand that embodiments of the present invention can be provided as methods, systems, or computer program products. Therefore, the present invention can take the form of a completely hardware embodiment, a completely software embodiment, or an embodiment combining software and hardware aspects. Furthermore, the present invention can take the form of a computer program product embodied on one or more computer-usable storage media (including, but not limited to, disk storage, CD-ROM, optical storage, etc.) containing computer-usable program code.

[0056] The above description is only a preferred embodiment of the present invention and is not intended to limit the present invention. Any modifications, equivalent substitutions, improvements, etc., made within the spirit and principles of the present invention should be included within the protection scope of the present invention.

Claims

1. A lightweight slope geological disaster monitoring and early warning method, characterized in that, include: S1. Standardize the pre-acquired real-time slope monitoring data and decompose the processed slope monitoring data to obtain a multi-scale influence layer. S2. Using the constructed third-order tensor coupling model, dynamic coupling calculations are performed on the multi-scale influence layers to obtain time-varying coupling parameters. The time-varying coupling parameters are then mapped to obtain a coupling matrix, which is used to reflect the degree of correlation between the multi-scale influence layers. S3. Using the standardized real-time slope monitoring data as a sample set, the isolated forest algorithm is used to calculate the risk factors from the real-time slope monitoring data. S4. The multi-scale influence layer, time-varying coupling strength parameters and risk factors are processed to obtain the input vector. The input vector is then calculated using a gated cyclic unit network to obtain the slope disaster risk index. S5. Based on the slope disaster risk index and combined with historical data of slope monitoring data, a warning threshold adjustment mechanism is constructed to adaptively adjust the warning threshold.

2. The lightweight slope geological disaster monitoring and early warning method according to claim 1, characterized in that, The process of standardizing the pre-acquired real-time slope monitoring data and decomposing the processed slope monitoring data to obtain a multi-scale influence layer includes: S11. Real-time slope monitoring data includes surface displacement data, soil tilt data, water pressure data, and rainfall data; S12. The real-time slope monitoring data is processed using the Z-Score standardization method to obtain a unified time-series dataset; S13. Perform time-frequency decomposition on the unified time series dataset to obtain a multi-scale influence layer, which includes a slow layer, a medium-speed layer, and a fast layer.

3. The lightweight slope geological disaster monitoring and early warning method according to claim 1, characterized in that, The process of using the constructed third-order tensor coupling model to perform dynamic coupling calculations on the multi-scale influence layer, obtaining time-varying coupling parameters, and mapping these parameters to obtain the coupling matrix includes: S21. Construct a time window using a preset time length, obtain vector data of the multi-scale influence layer in the same time window, and use the vector data as the calculation benchmark. S22. Within the time window, for any two influence layers in the multi-scale influence layer, calculate the covariance of their vector data and obtain the covariance value. S23. Based on the preset attenuation coefficient, the time-varying coupling parameters between the corresponding influence layers are calculated by combining the covariance value.

4. The lightweight slope geological disaster monitoring and early warning method according to claim 3, characterized in that, Within the time window, for any two influence layers in the multi-scale influence layers, the covariance of their vector data is calculated, and the covariance value includes: When the covariance is positive, any two influencing layers have a positive cooperative relationship; when the covariance is negative, any two influencing layers have a negative cooperative relationship.

5. The lightweight slope geological disaster monitoring and early warning method according to claim 1, characterized in that, The standardized real-time slope monitoring data is used as a sample set, and the isolated forest algorithm is used to calculate the risk factors, including: S31. Obtain standardized slope detection data, construct a historical sample set, use the historical sample set to construct an isolation tree, and extract multiple feature dimensions for each sample point in the historical sample set. Among these, several characteristic latitudes include displacement, tilt, water pressure, and rainfall; S32. Obtain the maximum and minimum values ​​of at least one feature dimension of the sample point, determine a split value between the maximum and minimum values, divide the sample point into left and right subtrees according to the split value, and repeat the division process until each sample point is isolated to a preset value. S33. Select at least one sample point from the historical sample set, and use the average length of the sample point to each isolation tree as the baseline length. S34. Acquire the sample to be tested in real time, traverse each isolation tree for the sample to be tested, and calculate the average length of the sample point in each isolation tree. S35. Calculate the average length of the sample points in the sample to be tested and the baseline length to obtain the probability value of abnormal sample points, and use this probability value as a risk factor.

6. The lightweight slope geological disaster monitoring and early warning method according to claim 1, characterized in that, The process involves processing multi-scale influence layers, time-varying coupling strength parameters, and risk factors to obtain an input vector. This input vector is then used to calculate the slope disaster risk index using a gated recurrent unit network. The resulting index includes: S41. The multi-scale influence layer, time-varying coupling strength parameters, and risk factors are concatenated to form the input vector at the current moment. S42. Input the current input vector into the gated recurrent unit network, and calculate the candidate hidden state and the hidden state through the reset gate and the update gate; S43. Calculate the hidden state using the fully connected layer to obtain the slope disaster risk index.

7. The lightweight slope geological disaster monitoring and early warning method according to claim 6, characterized in that, The step of inputting the current input vector into the gated recurrent unit network and calculating the candidate hidden state and the hidden state through the reset gate and update gate includes: S421. The previous hidden state is processed element-wise using the reset gate. The processed previous hidden state is concatenated with the current input vector to obtain a fused feature vector. The fused feature vector is then non-linearly mapped using an activation function to obtain the candidate hidden state. S422. The hidden state and candidate hidden state at the previous time step of the update gate are weighted and processed, and the element-wise weighted summation is used to calculate the result.

8. The lightweight slope geological disaster monitoring and early warning method according to claim 2, characterized in that, The expression for the multi-scale influence layer is: ； ； ； In the formula, Indicates a fast layer. Indicates medium speed layer ， Indicates a slow layer. G Represents surface displacement data. I Indicates soil tilt data, Represents rainfall data, Represents water pressure data. Indicates the wavelet scale. Represented as mother wavelet, Indicates the time shift amount. This indicates the rainfall over 24 hours. , and Both represent attenuation coefficients. This represents the change in water pressure over 24 hours. Indicates the long-term trend term. Indicates periodic fluctuations. This represents random noise and / or residuals.

9. A lightweight slope geological disaster monitoring and early warning method according to claim 3, characterized in that, The expression for the time-varying coupling parameter is: ； In the formula, and This represents the multi-scale influence layer, with values ​​ranging from [value range missing]. , ， , t Indicates a time window. This represents the impact layer data within the time window. Indicates the time window and The covariance operator, Indicates the time window The mean, Indicates the time window The mean, This represents the tensor decay factor between different influence layers. This represents the attenuation coefficient between multi-scale influence layers.

10. A lightweight slope geological disaster monitoring and early warning system, used to implement the lightweight slope geological disaster monitoring and early warning method according to any one of claims 1-9, characterized in that, include: The scale decomposition module is used to standardize the pre-acquired real-time slope monitoring data and decompose the processed slope monitoring data to obtain multi-scale influence layers. The coupling calculation module is used to utilize the constructed third-order tensor coupling model to perform dynamic coupling calculations on the multi-scale influence layers, obtain time-varying coupling parameters, and map the time-varying coupling parameters to obtain a coupling matrix, which is used to reflect the degree of correlation between the multi-scale influence layers. The risk calculation module is used to take the standardized real-time slope monitoring data as a sample set and use the isolated forest algorithm to calculate the risk factors from the real-time slope monitoring data. The index calculation module is used to process the multi-scale influence layer, time-varying coupling strength parameters and risk factors to obtain the input vector. The input vector is then calculated using a gated cyclic unit network to obtain the slope disaster risk index. The adaptive adjustment module is used to construct an early warning threshold adjustment mechanism based on the slope disaster risk index and combined with historical data of slope monitoring data, and to adaptively adjust the early warning threshold.

Images